US Census Geographic Boundaries

Cloud-optimized formats for Draft CalEnviroScreen 5.0 (released January 2026) is a cumulative environmental pollution and health vulnerability screening tool developed by California's Office of Environmental Health Hazard Assessment (OEHHA).

Product Details

Visibility: Public
Owner: Carl Boettiger
Created: 16 Jun 2026
Last Updated: 16 Jun 2026

Product Contents

root

census-2024

README-tract.md

tract-README.md

Product Details

Product Contents

US Census Geographic Boundaries

Product Details

Visibility: Public
Owner: Carl Boettiger
Created: 16 Jun 2026
Last Updated: 16 Jun 2026

Product Contents

root

census-2024

README-tract.md

tract-README.md

Name: tract-README.md
Size: 7.07 KB
Content Type: text/markdown; charset=utf-8
Last Modified: 16 Jun 2026
Source URL: https://data.source.coop/cboettig/census/census-2024/README-tract.md/tract-README.md
Cloud URI: s3://us-west-2.opendata.source.coop/cboettig/census/census-2024/README-tract.md/tract-README.md

Download

Name: tract-README.md
Size: 7.07 KB
Content Type: text/markdown; charset=utf-8
Last Modified: 16 Jun 2026
Source URL: https://data.source.coop/cboettig/census/census-2024/README-tract.md/tract-README.md
Cloud URI: s3://us-west-2.opendata.source.coop/cboettig/census/census-2024/README-tract.md/tract-README.md

Download

US Census 2024 Census Tracts

2024 census tract boundaries from the US Census Bureau's TIGER/Line shapefiles, processed into cloud-native formats with H3 hexagonal spatial indexing.

Dataset Overview

Source: US Census Bureau TIGER/Line 2024
Geography Level: Census Tract
Features: ~85,000 census tracts
H3 Resolution: 10 (~0.015 km² per hex)
Parent Resolutions: 9, 8, 0
Last Updated: February 2026

Available Formats

GeoParquet

s3://public-census/census-2024/tract.parquet

Format: GeoParquet with ZSTD compression
CRS: EPSG:4326 (WGS84)
Row Groups: 100,000 features
Geometry Type: MultiPolygon
Use Case: Analytical queries with DuckDB, Polars, GeoPandas

PMTiles

s3://public-census/census-2024/tract.pmtiles

Format: PMTiles v3
Use Case: Web map visualization with MapLibre GL JS, Leaflet

H3 Hex Index

s3://public-census/census-2024/tract/hex/h0={cell}/

Format: Partitioned Parquet by H3 resolution 0 cells
Columns: h3_cell (resolution 10), feature_id, plus parent resolutions
Use Case: Fast spatial joins, area-weighted aggregation

Attributes

Key attributes from TIGER/Line shapefiles:

STATEFP: State FIPS code (2 digits)
COUNTYFP: County FIPS code (3 digits)
TRACTCE: Census tract code (6 digits)
GEOID: Complete geographic identifier (STATEFP + COUNTYFP + TRACTCE)
NAME: Census tract name/number
NAMELSAD: Full name with legal/statistical description
ALAND: Land area (square meters)
AWATER: Water area (square meters)

Usage Examples

DuckDB Query

Python (GeoPandas)

Spatial Join Example

H3 Spatial Indexing

Each census tract is indexed into H3 hexagons at resolution 10 (~0.015 km² per hex):

Urban tracts: ~1,000-5,000 hexes (dense, small area)
Suburban tracts: ~5,000-20,000 hexes
Rural tracts: Can exceed 50,000 hexes for large areas

The hex index enables:

Fast point-in-tract queries: Convert points to H3, join with hex index (millions of points/second)
Area-weighted aggregation: Aggregate raster data to tracts using hex coverage
Multi-resolution analysis: Parent resolutions (9, 8, 0) support hierarchical queries

Census Tract Geography

Census tracts are small, relatively permanent statistical subdivisions of counties:

Typical population: 1,200 to 8,000 people (optimally 4,000)
Coverage: Entire US including territories
Designed for: Demographic data collection and analysis
Updates: Every 10 years with decennial census, boundaries may change

Processing Details

Preprocessing

Source Files: 56 per-state zip files
Method: Parallel download + unzip + merge via cng-convert-to-parquet
Duration: 3 minutes 2 seconds
Features: ~85,000 census tracts merged into single GeoParquet

Pipeline

Hex Generation: 200 chunks, 200/200 completed (10 minutes)
PMTiles: 18 minutes
Repartition: 2 minutes 14 seconds
Total Processing Time: ~23 minutes

Citation

U.S. Census Bureau (2024). TIGER/Line Shapefiles: Census Tracts. Retrieved from https://www.census.gov/geographies/mapping-files/time-series/geo/tiger-line-file.html

License

Public domain. No restrictions on use.

Census 2024 States - 56 state/territory boundaries
Census 2024 Counties - 3,235 county boundaries
Census 2024 Congressional Districts - 119th Congress districts

Access

Public HTTP Access:

S3 Access (requires credentials):

Notes

This dataset was preprocessed from 56 per-state zip files using parallel download and direct shapefile merging (see AGENTS.md for methodology). The efficient preprocessing approach reduced processing time from 30+ minutes (sequential) to ~3 minutes (parallel).

US Census 2024 Census Tracts

2024 census tract boundaries from the US Census Bureau's TIGER/Line shapefiles, processed into cloud-native formats with H3 hexagonal spatial indexing.

Dataset Overview

Source: US Census Bureau TIGER/Line 2024
Geography Level: Census Tract
Features: ~85,000 census tracts
H3 Resolution: 10 (~0.015 km² per hex)
Parent Resolutions: 9, 8, 0
Last Updated: February 2026

Available Formats

GeoParquet

s3://public-census/census-2024/tract.parquet

Format: GeoParquet with ZSTD compression
CRS: EPSG:4326 (WGS84)
Row Groups: 100,000 features
Geometry Type: MultiPolygon
Use Case: Analytical queries with DuckDB, Polars, GeoPandas

PMTiles

s3://public-census/census-2024/tract.pmtiles

Format: PMTiles v3
Use Case: Web map visualization with MapLibre GL JS, Leaflet

H3 Hex Index

s3://public-census/census-2024/tract/hex/h0={cell}/

Format: Partitioned Parquet by H3 resolution 0 cells
Columns: h3_cell (resolution 10), feature_id, plus parent resolutions
Use Case: Fast spatial joins, area-weighted aggregation

Attributes

Key attributes from TIGER/Line shapefiles:

STATEFP: State FIPS code (2 digits)
COUNTYFP: County FIPS code (3 digits)
TRACTCE: Census tract code (6 digits)
GEOID: Complete geographic identifier (STATEFP + COUNTYFP + TRACTCE)
NAME: Census tract name/number
NAMELSAD: Full name with legal/statistical description
ALAND: Land area (square meters)
AWATER: Water area (square meters)

Usage Examples

DuckDB Query

Python (GeoPandas)

Spatial Join Example

H3 Spatial Indexing

Each census tract is indexed into H3 hexagons at resolution 10 (~0.015 km² per hex):

Urban tracts: ~1,000-5,000 hexes (dense, small area)
Suburban tracts: ~5,000-20,000 hexes
Rural tracts: Can exceed 50,000 hexes for large areas

The hex index enables:

Fast point-in-tract queries: Convert points to H3, join with hex index (millions of points/second)
Area-weighted aggregation: Aggregate raster data to tracts using hex coverage
Multi-resolution analysis: Parent resolutions (9, 8, 0) support hierarchical queries

Census Tract Geography

Census tracts are small, relatively permanent statistical subdivisions of counties:

Typical population: 1,200 to 8,000 people (optimally 4,000)
Coverage: Entire US including territories
Designed for: Demographic data collection and analysis
Updates: Every 10 years with decennial census, boundaries may change

Processing Details

Preprocessing

Source Files: 56 per-state zip files
Method: Parallel download + unzip + merge via cng-convert-to-parquet
Duration: 3 minutes 2 seconds
Features: ~85,000 census tracts merged into single GeoParquet

Pipeline

Hex Generation: 200 chunks, 200/200 completed (10 minutes)
PMTiles: 18 minutes
Repartition: 2 minutes 14 seconds
Total Processing Time: ~23 minutes

Citation

U.S. Census Bureau (2024). TIGER/Line Shapefiles: Census Tracts. Retrieved from https://www.census.gov/geographies/mapping-files/time-series/geo/tiger-line-file.html

License

Public domain. No restrictions on use.

Census 2024 States - 56 state/territory boundaries
Census 2024 Counties - 3,235 county boundaries
Census 2024 Congressional Districts - 119th Congress districts

Access

Public HTTP Access:

S3 Access (requires credentials):

Notes

1https://s3-west.nrp-nautilus.io/public-census/census-2024/tract.parquet
2https://s3-west.nrp-nautilus.io/public-census/census-2024/tract.pmtiles

1s3://public-census/census-2024/tract.parquet
2s3://public-census/census-2024/tract.pmtiles
3s3://public-census/census-2024/tract/hex/h0=*/

1https://s3-west.nrp-nautilus.io/public-census/census-2024/tract.parquet
2https://s3-west.nrp-nautilus.io/public-census/census-2024/tract.pmtiles

1s3://public-census/census-2024/tract.parquet
2s3://public-census/census-2024/tract.pmtiles
3s3://public-census/census-2024/tract/hex/h0=*/

1-- Load census tract boundaries
2LOAD spatial;
3CREATE TABLE tracts AS 
4  SELECT * FROM read_parquet('s3://public-census/census-2024/tract.parquet');
5
6-- Query tracts in a specific county (Los Angeles)
7SELECT NAME, ALAND/1000000 as land_sq_km
8FROM tracts
9WHERE STATEFP = '06' AND COUNTYFP = '037'
10ORDER BY NAME;
11
12-- Find tracts with minimal land area (often islands or special cases)
13SELECT GEOID, NAME, ALAND
14FROM tracts
15WHERE ALAND < 100000
16ORDER BY ALAND;
17
18-- Join with H3 hex index for spatial analysis
19CREATE TABLE tract_hexes AS
20  SELECT * FROM read_parquet('s3://public-census/census-2024/tract/hex/h0=*/');
21  
22-- Calculate hex density per tract
23SELECT t.GEOID, t.NAME, 
24       COUNT(h.h3_cell) as hex_count,
25       t.ALAND/1000000 as land_sq_km,
26       COUNT(h.h3_cell) / (t.ALAND/1000000) as hexes_per_sq_km
27FROM tract_hexes h
28JOIN tracts t ON h.feature_id = t._cng_fid
29WHERE t.STATEFP = '06' AND t.COUNTYFP = '037'
30GROUP BY t.GEOID, t.NAME, t.ALAND
31ORDER BY hex_count DESC
32LIMIT 20;

1-- Load census tract boundaries
2LOAD spatial;
3CREATE TABLE tracts AS 
4  SELECT * FROM read_parquet('s3://public-census/census-2024/tract.parquet');
5
6-- Query tracts in a specific county (Los Angeles)
7SELECT NAME, ALAND/1000000 as land_sq_km
8FROM tracts
9WHERE STATEFP = '06' AND COUNTYFP = '037'
10ORDER BY NAME;
11
12-- Find tracts with minimal land area (often islands or special cases)
13SELECT GEOID, NAME, ALAND
14FROM tracts
15WHERE ALAND < 100000
16ORDER BY ALAND;
17
18-- Join with H3 hex index for spatial analysis
19CREATE TABLE tract_hexes AS
20  SELECT * FROM read_parquet('s3://public-census/census-2024/tract/hex/h0=*/');
21  
22-- Calculate hex density per tract
23SELECT t.GEOID, t.NAME, 
24       COUNT(h.h3_cell) as hex_count,
25       t.ALAND/1000000 as land_sq_km,
26       COUNT(h.h3_cell) / (t.ALAND/1000000) as hexes_per_sq_km
27FROM tract_hexes h
28JOIN tracts t ON h.feature_id = t._cng_fid
29WHERE t.STATEFP = '06' AND t.COUNTYFP = '037'
30GROUP BY t.GEOID, t.NAME, t.ALAND
31ORDER BY hex_count DESC
32LIMIT 20;

1import geopandas as gpd
2import pandas as pd
3
4# Read census tracts
5tracts = gpd.read_parquet('s3://public-census/census-2024/tract.parquet')
6
7# Filter to a specific state (California)
8ca_tracts = tracts[tracts['STATEFP'] == '06']
9print(f"California census tracts: {len(ca_tracts)}")
10
11# Urban vs rural tracts (using land area as proxy)
12ca_tracts['area_km2'] = ca_tracts['ALAND'] / 1e6
13urban_threshold = 10  # sq km
14urban = ca_tracts[ca_tracts['area_km2'] < urban_threshold]
15rural = ca_tracts[ca_tracts['area_km2'] >= urban_threshold]
16
17print(f"Urban tracts: {len(urban)} ({len(urban)/len(ca_tracts)*100:.1f}%)")
18print(f"Rural tracts: {len(rural)} ({len(rural)/len(ca_tracts)*100:.1f}%)")
19
20# Find coastal tracts (high water percentage)
21ca_tracts['water_pct'] = (ca_tracts['AWATER'] / 
22                           (ca_tracts['ALAND'] + ca_tracts['AWATER'])) * 100
23coastal = ca_tracts[ca_tracts['water_pct'] > 20].sort_values('water_pct', ascending=False)
24print(coastal[['GEOID', 'NAME', 'water_pct']].head())

1import geopandas as gpd
2import pandas as pd
3
4# Read census tracts
5tracts = gpd.read_parquet('s3://public-census/census-2024/tract.parquet')
6
7# Filter to a specific state (California)
8ca_tracts = tracts[tracts['STATEFP'] == '06']
9print(f"California census tracts: {len(ca_tracts)}")
10
11# Urban vs rural tracts (using land area as proxy)
12ca_tracts['area_km2'] = ca_tracts['ALAND'] / 1e6
13urban_threshold = 10  # sq km
14urban = ca_tracts[ca_tracts['area_km2'] < urban_threshold]
15rural = ca_tracts[ca_tracts['area_km2'] >= urban_threshold]
16
17print(f"Urban tracts: {len(urban)} ({len(urban)/len(ca_tracts)*100:.1f}%)")
18print(f"Rural tracts: {len(rural)} ({len(rural)/len(ca_tracts)*100:.1f}%)")
19
20# Find coastal tracts (high water percentage)
21ca_tracts['water_pct'] = (ca_tracts['AWATER'] / 
22                           (ca_tracts['ALAND'] + ca_tracts['AWATER'])) * 100
23coastal = ca_tracts[ca_tracts['water_pct'] > 20].sort_values('water_pct', ascending=False)
24print(coastal[['GEOID', 'NAME', 'water_pct']].head())

1import duckdb
2
3# Point-in-tract assignment using H3 join
4con = duckdb.connect()
5con.execute("LOAD spatial")
6
7# Your point data with lat/lon
8points_df = pd.DataFrame({
9    'event_id': range(1000),
10    'lat': [...],  # your latitudes
11    'lon': [...]   # your longitudes
12})
13
14# Convert points to H3 cells (resolution 10)
15con.execute("""
16    INSTALL h3 FROM community;
17    LOAD h3;
18    
19    CREATE TABLE points AS
20    SELECT event_id, h3_latlng_to_cell(lat, lon, 10) as h3_cell
21    FROM points_df;
22    
23    -- Join with tract hex index
24    CREATE TABLE events_by_tract AS
25    SELECT t.GEOID, t.NAME, COUNT(*) as event_count
26    FROM points p
27    JOIN read_parquet('s3://public-census/census-2024/tract/hex/h0=*/') h 
28      ON p.h3_cell = h.h3_cell
29    JOIN read_parquet('s3://public-census/census-2024/tract.parquet') t 
30      ON h.feature_id = t._cng_fid
31    GROUP BY t.GEOID, t.NAME
32    ORDER BY event_count DESC;
33""")
34
35results = con.execute("SELECT * FROM events_by_tract").df()

1import duckdb
2
3# Point-in-tract assignment using H3 join
4con = duckdb.connect()
5con.execute("LOAD spatial")
6
7# Your point data with lat/lon
8points_df = pd.DataFrame({
9    'event_id': range(1000),
10    'lat': [...],  # your latitudes
11    'lon': [...]   # your longitudes
12})
13
14# Convert points to H3 cells (resolution 10)
15con.execute("""
16    INSTALL h3 FROM community;
17    LOAD h3;
18    
19    CREATE TABLE points AS
20    SELECT event_id, h3_latlng_to_cell(lat, lon, 10) as h3_cell
21    FROM points_df;
22    
23    -- Join with tract hex index
24    CREATE TABLE events_by_tract AS
25    SELECT t.GEOID, t.NAME, COUNT(*) as event_count
26    FROM points p
27    JOIN read_parquet('s3://public-census/census-2024/tract/hex/h0=*/') h 
28      ON p.h3_cell = h.h3_cell
29    JOIN read_parquet('s3://public-census/census-2024/tract.parquet') t 
30      ON h.feature_id = t._cng_fid
31    GROUP BY t.GEOID, t.NAME
32    ORDER BY event_count DESC;
33""")
34
35results = con.execute("SELECT * FROM events_by_tract").df()

US Census Geographic Boundaries

US Census Geographic Boundaries

US Census 2024 Census Tracts

Dataset Overview

Available Formats

GeoParquet

PMTiles

H3 Hex Index

Attributes

Usage Examples

DuckDB Query

Python (GeoPandas)

Spatial Join Example

H3 Spatial Indexing

Census Tract Geography

Processing Details

Preprocessing

Pipeline

Citation

License

Related Datasets

Access

Notes

US Census 2024 Census Tracts

Dataset Overview

Available Formats

GeoParquet

PMTiles

H3 Hex Index

Attributes

Usage Examples

DuckDB Query

Python (GeoPandas)

Spatial Join Example

H3 Spatial Indexing

Census Tract Geography

Processing Details

Preprocessing

Pipeline

Citation

License

Related Datasets

Access

Notes